An Application of SVD++ Method in Collaborative Filtering

Collaborative filtering algorithms have important applications in the implementation of recommendation systems. Collaborative filtering generally focuses on the user's evaluation scores for commodity items, and these evaluation data imply some specific relevance. Therefore, this type of collaborative filtering Model research has attracted widespread attention. SVD is currently one of the most classic and practical collaborative filtering algorithm models. It can deal with the problem of data sparseness in the recommendation system and derive different types of SVD models on this basis, but its recommendation results It has never been significantly improved. In response to such problems, this article proposes a new variant of the SVD++ algorithm that incorporates a special timing mechanism for dynamic adjustment, and uses the average absolute error, root mean square error and standard average absolute error for the recommended results. After evaluation, we got better results than the current classic SVD++ model.